ResNet-2Plus1D: Optimized for Qualcomm Devices

ResNet (2+1)D Convolutions is a network which explicitly factorizes 3D convolution into two separate and successive operations, a 2D spatial convolution and a 1D temporal convolution. It used for video understanding applications.

This is based on the implementation of ResNet-2Plus1D found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Getting Started

There are two ways to deploy this model on your device:

Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

Runtime	Precision	Chipset	SDK Versions	Download
ONNX	float	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit ResNet-2Plus1D on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

• Custom weights (e.g., fine-tuned checkpoints)
• Custom input shapes
• Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for ResNet-2Plus1D on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.video_classification

Model Stats:

• Model checkpoint: Kinetics-400
• Input resolution: 112x112
• Number of parameters: 31.5M
• Model size (float): 120 MB
• Model size (w8a8): 30.8 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	5.523 ms	2 - 215 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® X2 Elite	6.149 ms	60 - 60 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® X Elite	12.164 ms	60 - 60 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.048 ms	0 - 290 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS8550 (Proxy)	11.987 ms	2 - 5 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS9075	21.351 ms	2 - 7 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.219 ms	0 - 209 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.923 ms	0 - 191 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X2 Elite	2.048 ms	31 - 31 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X Elite	4.542 ms	31 - 31 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	3.263 ms	0 - 228 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS6490	324.446 ms	97 - 127 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	4.322 ms	0 - 5 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS9075	4.112 ms	1 - 3 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCM6690	302.049 ms	100 - 108 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.594 ms	0 - 189 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	263.662 ms	68 - 76 MB	CPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.641 ms	2 - 227 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X2 Elite	6.693 ms	2 - 2 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X Elite	12.887 ms	2 - 2 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.229 ms	0 - 291 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	81.889 ms	1 - 214 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	12.624 ms	2 - 5 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8775P	21.268 ms	0 - 214 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS9075	23.064 ms	4 - 8 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	29.099 ms	0 - 270 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA7255P	81.889 ms	1 - 214 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8295P	22.733 ms	0 - 197 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.175 ms	0 - 216 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.845 ms	1 - 184 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X2 Elite	2.32 ms	1 - 1 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X Elite	4.909 ms	1 - 1 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.351 ms	0 - 221 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS6490	19.543 ms	3 - 5 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	13.405 ms	1 - 182 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.616 ms	0 - 10 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.664 ms	1 - 186 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS9075	4.763 ms	1 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCM6690	84.546 ms	1 - 199 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	7.751 ms	1 - 217 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA7255P	13.405 ms	1 - 182 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8295P	7.848 ms	1 - 182 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.575 ms	1 - 182 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.885 ms	1 - 188 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	258.531 ms	0 - 247 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	297.22 ms	0 - 316 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8275 (Proxy)	714.79 ms	0 - 237 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8550 (Proxy)	377.527 ms	0 - 3 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8775P	392.608 ms	0 - 236 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS9075	388.273 ms	0 - 66 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8450 (Proxy)	428.651 ms	0 - 301 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA7255P	714.79 ms	0 - 237 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8295P	452.308 ms	0 - 227 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	262.417 ms	0 - 231 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	643.194 ms	0 - 455 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	578.421 ms	0 - 514 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS6490	1786.683 ms	267 - 428 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	1520.257 ms	0 - 439 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	755.68 ms	0 - 3 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8775P	798.925 ms	0 - 439 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS9075	572.04 ms	0 - 65 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCM6690	1606.748 ms	310 - 471 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	862.725 ms	0 - 466 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA7255P	1520.257 ms	0 - 439 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8295P	869.513 ms	0 - 435 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	508.65 ms	0 - 546 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1224.497 ms	294 - 365 MB	NPU

License

• The license for the original implementation of ResNet-2Plus1D can be found here.

References

• A Closer Look at Spatiotemporal Convolutions for Action Recognition
• Source Model Implementation

Community

• Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
• For questions or feedback please reach out to us.